Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(6.5), p. 47 — X=QX~QTX=Q\tilde XQ^TX=QX~QT is positive definite and satisfies ρX−X−1=ρ(Zk−Uk)−S\rho X-X^{-1}=\rho(Z^k-U^k)-SρX−X−1=ρ(Zk−Uk)−S

Proved
BoydADMM.L1.covsel_eq_6_5

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

admmcovariance-selectioneigenvalue-decompositionp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-bookp2o-v1

Let S,Z,U∈Rn×nS,Z,U\in\mathbb R^{n\times n}S,Z,U∈Rn×n, ρ>0\rho>0ρ>0, let Q∈Rn×nQ\in\mathbb R^{n\times n}Q∈Rn×n satisfy QTQ=QQT=IQ^TQ=QQ^T=IQTQ=QQT=I, and let μ1,…,μn∈R\mu_1,\dots,\mu_n\in\mathbb Rμ1​,…,μn​∈R be such that

ρ(Z−U)−S=Q Λ QT,Λ=diag(μ1,…,μn).\rho(Z-U)-S=Q\,\Lambda\,Q^T,\qquad \Lambda=\mathrm{diag}(\mu_1,\dots,\mu_n).ρ(Z−U)−S=QΛQT,Λ=diag(μ1​,…,μn​).

Put X~=diag(X~11,…,X~nn)\tilde X=\mathrm{diag}(\tilde X_{11},\dots,\tilde X_{nn})X~=diag(X~11​,…,X~nn​) with X~ii=(μi+μi2+4ρ)/(2ρ)\tilde X_{ii}=(\mu_i+\sqrt{\mu_i^2+4\rho})/(2\rho)X~ii​=(μi​+μi2​+4ρ​)/(2ρ), and X=QX~QTX=Q\tilde XQ^TX=QX~QT. Then XXX is positive definite and

ρX−X−1=ρ(Z−U)−S.(6.5)\rho X-X^{-1}=\rho(Z-U)-S. \tag{6.5}ρX−X−1=ρ(Z−U)−S.(6.5)

Equation (6.5) is the first-order optimality condition of the X-update in ADMM for sparse inverse covariance selection, so this says that the book's eigenvalue construction produces a positive definite solution of it.

Formalization Note Z,UZ,UZ,U stand for Zk,UkZ^k,U^kZk,Uk; the eigenvalues λi\lambda_iλi​ are called μi\mu_iμi​. The statement quantifies over every orthogonal QQQ and diagonal Λ\LambdaΛ with ρ(Z−U)−S=QΛQT\rho(Z-U)-S=Q\Lambda Q^Tρ(Z−U)−S=QΛQT, not over one decomposition chosen by Mathlib; such a decomposition exists whenever ρ(Z−U)−S\rho(Z-U)-Sρ(Z−U)−S is symmetric, and the identity forces it to be symmetric, so no symmetry hypothesis on SSS or Z−UZ-UZ−U is needed here. X−1X^{-1}X−1 is Mathlib's matrix inverse, the true inverse since XXX is positive definite.

Preamble
import Mathlib
import Definitions.Def_BoydADMM_L1_CovSel

open Matrix
Formal statement
namespace BoydADMM.L1

/-- §6.5, (6.5), p. 47: if `ρ > 0`, `Q` is orthogonal (`QᵀQ = QQᵀ = I`) and
`ρ(Z − U) − S = Q diag(μ) Qᵀ`, then `X = Q X̃ Qᵀ` with `X̃_ii = (μ_i + √(μ_i² + 4ρ))/(2ρ)` is
positive definite and satisfies `ρX − X⁻¹ = ρ(Z − U) − S`. -/
theorem covsel_eq_6_5 {n : ℕ} (S Z U Q : Matrix (Fin n) (Fin n) ℝ) (ρ : ℝ) (μ : Fin n → ℝ)
    (hρ : 0 < ρ) (hQ₁ : Qᵀ * Q = 1) (hQ₂ : Q * Qᵀ = 1)
    (hdecomp : ρ • (Z - U) - S = Q * Matrix.diagonal μ * Qᵀ) :
    (covselX ρ Q μ).PosDef ∧ ρ • covselX ρ Q μ - (covselX ρ Q μ)⁻¹ = ρ • (Z - U) - S := by sorry

end BoydADMM.L1
Source
Boyd, Parikh, Chu, Peleato, Eckstein, Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers, Found. Trends Mach. Learn. 3(1) (2011), p. 47, §6.5, (6.5)
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me